Th sensitivity of 3D separations in multi-stage compressors

One of the most important questions facing a compressor designer is how accurately their computational code predicts the size of 3D corner separations. This is because the size of the 3D corner separation sets the compressors off-design performance and blockage and therefore it’s operating range. The aim of this paper is to determine how two types of modelling fidelities limit the accuracy of the prediction of 3D corner separations. The first modelling fidelity is the accuracy with which the flow within a blade passage is modelled (endwall geometry modelling, shroud modelling and the boundary layer transition modelling). The second modelling fidelity is the accuracy of modelling in the multi-blade row environment (multi-stage repeating stage boundary layer modelling and the local skew endwall boundary layer modelling). The first part of the paper shows that when the compressor studied has its original “design intent” geometry (defined as 1% shroud clearance and a hydraulically smooth blade surface) the size of the 3D corner separation is relatively insensitive to the modelling fidelities within the blade row but is highly sensitive to modelling fidelity in the multi-blade row environment. This explains the inability of 3D CFD to predict multistage matching. The second part of the paper shows that when the compressor has its “end of life” geometry (defined as 3% shroud clearance and surface roughness measured from an engine blade at 4,000 cycles) the size of the 3D corner separation suddenly becomes highly sensitivity to modelling fidelities within an individual blade row. This finding is of significant importance to designers because it shows that current computational codes are not able to predict end of life performance